The metabotropic glutamate receptor 5 radioligand [11C]AZD9272 identifies unique binding sites in primate brain

&NA; The metabotropic glutamate receptor 5 (mGluR5) is a target for drug development and for imaging studies of the glutamate system in neurological and psychiatric disorders. [11C]AZD9272 is a selective mGluR5 PET radioligand that is structurally different from hitherto applied mGluR5 radioligands. In the present investigation we compared the binding patterns of radiolabeled AZD9272 and other mGluR5 radioligands in the non‐human primate (NHP) brain. PET studies were undertaken using [11C]AZD9272 and the commonly applied mGluR5 radioligand [11C]ABP688. Autoradiography studies were performed in vitro using [3H]AZD9272 and the standard mGluR5 radioligands [3H]M‐MTEP and [3H]ABP688 in NHP tissue. Competition binding studies were undertaken in vivo and in vitro using different mGluR5 selective compounds as inhibitors. In comparison to other mGluR5 radioligands radiolabeled AZD9272 displayed a distinct regional distribution pattern with high binding in ventral striatum, midbrain, thalamus and cerebellum. While the binding of [11C]AZD9272 was almost completely inhibited by the structurally unique mGluR5 compound fenobam (2.0 mg/kg; 98% occupancy), it was only partially inhibited (46% and 20%, respectively) by the mGluR5 selective compounds ABP688 and MTEP, at a dose (2.0 mg/kg) expected to saturate the mGluR5. Autoradiography studies using [3H]AZD9272 confirmed a distinct pharmacologic profile characterized by preferential sensitivity to fenobam. The distinctive binding in ventral striato‐pallido‐thalamic circuits and shared pharmacologic profile with the pro‐psychotic compound fenobam warrants further examination of [11C]AZD9272 for potential application in psychiatric neuroimaging studies. HighlightsmGluR5 radioligand [11C]AZD9272 was characterized in the non‐human primate brain.Binding distribution of [11C]AZD9272 differs from that of other mGluR5 radioligands.Unique AZD9272 binding sites are found in ventral striatum, thalamus and midbrain.AZD9272 can be fully displaced from its binding sites by fenobam.mGluR5 specific MPEP‐like ligands induce only partial inhibition of AZD9272 binding.


Introduction
The glutamate system is the major excitatory neurotransmitter system in brain (Rothman et al., 2003). Glutamatergic signaling is mediated through ionotropic and metabotropic, G-protein coupled receptors (GPCRs; Niciu et al., 2012). Of the glutamate receptors, the metabotropic glutamate receptor 5 (mGluR5) has attracted particular interest as an imaging biomarker and a potential target for the pharmacologic treatment of a number of CNS disorders (Pillai and Tipre, 2016). Non-competitive mGluR5 agonists (positive allosteric modulators, PAMs) have been developed for the treatment of psychosis  and antagonists (negative allosteric modulators, NAMs) are under evaluation for the treatment of several conditions, including depression, anxiety, neuropathic pain and levodopa-induced dyskinesia (Emmitte, 2013).
Importantly, mGluR5 modulators bind to sites that are topographically separated from the site recognized by glutamate (Pagano et al., 2000), thereby circumventing problems of subtype selectivity inherent to the use of orthosteric ligands ). An allosteric binding site has been identified in the transmembrane domain of the mGluR5 and appears to be shared by different classes of mGluR5-active compounds (Chen et al., 2007;Malherbe et al., 2006). However, some more recently discovered mGluR5 PAMs have been found to interact at allosteric sites that are distinct from that of early reference mGluR5 compounds (Chen et al., 2008;Hammond et al., 2010;Noetzel et al., 2013;O'Brien et al., 2004). Interestingly, this difference has been demonstrated to translate into activation of different intracellular signaling pathways (Noetzel et al., 2013;Zhang et al., 2005). In addition, the mGluR5 may form heteromeric complexes with other GPCRs, which could further add to the diversity in binding properties of and functional responses to mGluR5 compounds in vivo (Canela et al., 2009;Fuxe et al., 2009;Schr€ oder et al., 2009).
A number of radioligands targeting the mGluR5 have been developed. Tritium-labeled compounds, including [ 3 H]methoxy-MTEP ([ 3 H]M-MTEP; Anderson et al., 2002) and [ 3 H]methoxy-MPEP (Gasparini et al., 2002) are suitable tools for evaluation of mGluR5 related pharmacology in vitro, whereas [ 11 C]ABP688 (Ametamey et al., 2007) and [ 18 F]FPEB (Sullivan et al., 2013;Wong et al., 2013) are the radioligands most widely used for PET studies in vivo. All of the frequently used radioligands are chemical analogues of the prototype mGluR5 NAM compound MPEP and have been shown to display similar binding characteristics (Anderson et al., 2002;Gasparini et al., 2002;Hintermann et al., 2007). In light of the diverse pharmacologic properties of mGluR5 compounds, characterization of radioligands with different structural features may provide further insight into mGluR5 pharmacology and add tools for imaging of mGluR5 binding in vivo.
[ 11 C]AZD9272 is an mGluR5 NAM radioligand developed in collaboration between AstraZeneca and Karolinska Institutet (Andersson et al., 2013;Kågedal et al., 2012). Unlike the more commonly applied radioligands, [ 11 C]AZD9272 has been derived from a family of compounds that is structurally unrelated to MPEP. AZD9272 has been found to be highly selective for mGluR5 over 134 other receptors, ion channels, transporters and enzymes  and to bind with high affinity to human recombinant (K D ¼ 2.8 nM) and native rat (K D ¼ 9.4 nM) mGluR5 . The binding of [ 3 H]AZD9272 in rat brain tissue has been reported to be inhibited by MPEP, consistent with a common binding site of AZD9272 and MPEP-like compounds . In addition, drug discriminative properties of AZD9272 in rats have been found to be shared with those of the selective mGluR5 NAM MTEP supporting similar behavioral pharmacology of AZD9272 and other mGluR5 antagonists in rodents (Swedberg and Raboisson, 2014).
The compound has been evaluated in clinical trials for the treatment of neuropathic pain, but was, however, withdrawn at early stages of development due to clinical observations of psychotic symptoms (Ståhle et al., 2012). Similar psychotomimetic effects have been reported with fenobam (Friedmann et al., 1980;Pecknold et al., 1982), an anxiolytic compound developed in the 1970s, and later found to be an mGluR5 NAM (Porter et al., 2005). These observations are in contrast to the experience with other mGluR5 NAM drug candidates that have shown more favorable safety profiles and proceeded to later stages of clinical development (Quiroz et al., 2016).
In the present study, the cerebral binding pattern of radiolabeled AZD9272 was compared with that of hitherto applied radioligands for mGluR5. The in vivo imaging characteristics of [ 11 C] AZD9272 and [ 11 C]ABP688 were compared using PET studies in non-human primate (NHP), and the binding profile of [ 11 C] AZD9272 was assessed in occupancy studies comparing the binding at baseline and after administration of fenobam, ABP688 or MTEP. To provide high-resolution correlates to the PET images, autoradiography, using [ 3 H]AZD9272 and [ 3 H]M-MTEP as radioligands, was performed in NHP brain tissue. In addition, competition binding studies using autoradiography and [ 3 H]AZD9272 and [ 3 H]M-MTEP as radioligands were undertaken at increasing concentrations of fenobam, ABP688, or the more recently developed mGluR5 NAM AZD2066 (Swedberg and Raboisson, 2014).
The study was approved by the Animal Ethics Committee of the Swedish Animal Welfare Agency (Dnr 145/08, 399/08 and 386/09) and was performed according to the "Guidelines for planning, conducting and documenting experimental research" (Dnr 4820/ 06e600) of the Karolinska Institutet and the "Guide for the Care and Use of Laboratory Animals" (Clark et al., 1997 with an esophageal thermometer. Heart and respiration rates were continuously monitored during the experiment. No anesthesia or drug related effect on vital parameters or other functions measured (oxygen saturation, rectal temperature) were noted. The NHP head was immobilized in a head fixation system (Karlsson et al., 1993).
In each PET measurement a sterile physiological phosphate buffer solution (pH ¼ 7.4) of the radiotracer was injected as a bolus into a sural vein during 5 s simultaneously with the start of PET data acquisition. PET measurements were conducted using the High Resolution Research Tomograph (HRRT; Siemens Molecular Imaging, Knoxville, TN, USA). Emission data were acquired in list mode for 123 min. Dynamic images were reconstructed using threedimensional ordinary Poisson ordered subset expectation maximization including modeling of the system's point spread function. This procedure has previously been shown to result in a resolution of approximately 2 mm (Varrone et al., 2009). The radioactivity injected was 131e210 MBq for [ 11 C]AZD9272 and 191e211 MBq for Arterial blood was sampled as previously described (Finnema et al., 2014) using an automated blood sampling system during the first 3 min of each PET measurement. Subsequently, arterial blood samples (1e3 mL) were manually drawn at 3, 4, 5, 8, 15, 30, 45, 60, 75 and 90 min after injection. After centrifugation 0.2e1.5 mL plasma was pipetted and plasma radioactivity was measured in a well counter. In addition, samples were taken directly from the ABSS at 0.5, 1, 1.5, 2 and 2.5 min for crosscalibration with the well counter and for determination of the plasma to blood ratio.
The fraction of plasma radioactivity corresponding to the unchanged radioligand in plasma was determined as previously described (Halldin et al., 1995). Briefly, arterial plasma samples of 0.4e1.5 mL, sampled at 5, 15, 30, 45, 60, 75 and 90 min after injection, were deproteinized with acetonitrile and analyzed by gradient high-performance liquid chromatography with radiodetection. The unbound fractions of radioligands in plasma were analyzed using ultrafiltration as previously described (Amini et al., 2014).
Competition binding studies using [ 11 C]AZD9272 as a radioligand, and AZD9272, fenobam, ABP688 or MTEP (for description of the compounds, see Supplementary Table S1) as inhibitors, were carried out in four NHPs. In each experimental session two PET measurements were conducted, approximately 4 h apart, at baseline and after drug administration. Test compounds were administered as a 10 min intravenous infusion starting 15 min prior to the PET measurement. Doses administered were 0.4 mg/kg for AZD9272, 2.0 mg/kg for fenobam, and 0.25 mg/kg, 1.0 mg/kg and 2.0 mg/kg for both MTEP and ABP688. Doses represent the free base of the compounds. The dose selected for fenobam and the highest doses of MTEP administered were predicted to occupy more than 80% of the binding sites in brain based on previous findings using [ 11 C]ABP688 in baboon (DeLorenzo et al., 2011;Mathews et al., 2014).
PET measurements with [ 11 C]ABP688 were undertaken in three NHPs. The [ 11 C]ABP688 PET studies included baseline measurements in the two additional NHPs, and two consecutive PET measurements, at baseline and after administration of AZD9272 (0.4 mg/kg), respectively, in one of the four NHPs previously examined using [ 11 C]AZD9272. T1-weighted magnetic resonance images (MRIs) of the individual NHP brains had been previously obtained using a 1.5 T General Electrics Signa (GE, Milwakee, WI, USA) system. The MR sequence was a 3D spoiled gradient recalled protocol with the following settings: repetition time, 21 ms; flip angle, 35 ; field of view, 12.8 cm; matrix, 256 Â 256 Â 128; 128 Â 1.0 mm slices; and number of excitations, 2. Regions of interest (ROIs) for cortical regions (dorsal prefrontal (dPFC), anterior cingulate (ACC), insular (INS), posterior temporal (pTC) and occipital cortex (OC)), hippocampus (HIP), caudate nucleus (CAU), putamen (PUT), ventral striatum (VST), thalamus (THA), ventral midbrain (VM) and cerebellum (CER) were manually delineated on T1-weighted MRIs using an in-house image analysis software (Roland et al., 1994). MRIs were coregistered to averaged (0e123 min) PET images using SPM5 (Wellcome Department of Imaging Neuroscience, UK). Timeactivity curves were generated by pooling ROIs for each paired anatomical region and applying the pooled ROIs to PET images using the affine transformation matrix acquired from coregistration of the MRI.
PET data were analyzed using the software PMOD v. 3.6. Regional estimates of the total volume of distribution (V T ) for [ 11 C] ABP688 and [ 11 C]AZD9272 were obtained using the Logan linear graphical method (Logan et al., 1990). V T values were calculated based on 90 min of PET data acquisition. Occupancy at [ 11 C]ABP688 and [ 11 C]AZD9272 binding sites and non-displaceable volume of distribution (V ND ; Innis et al., 2007) were estimated based on regional V T values obtained at baseline (V T, baseline ) and after drug administration (V T, drug ) according to a graphical method described in the literature (Cunningham et al., 2010).
For NHPs examined with [ 11 C]AZD9272 individual values for V ND were calculated based on V T, baseline and V T, drug derived from experiments conducted using the highest dose of the four test compounds. V ND for [ 11 C]ABP688 was calculated as described above for the NHP examined at baseline and after administration of AZD9272. For the two additional NHPs the V ND for [ 11 C]ABP688 estimated as described above was normalized to individual V T values obtained for the cerebellum at baseline. Specific volumes of distribution (V S ; Innis et al., 2007) were subsequently calculated by subtracting individual estimates of V ND from regional V T values.

Radiosynthesis of [ 3 H]AZD9272, [ 3 H]M-MTEP and [ 3 H]ABP688
, and in vitro autoradiography studies were conducted at AstraZeneca, S€ odert€ alje, Sweden, in accordance with previously described procedures . Fresh frozen brain tissue from NHP was obtained under approved ethical guidelines and was processed as previously described (Pierson et al., 2008 Radioactivity was detected and quantified using a Fujifilm FLA7000 phosphor imager (Fuji, Tokyo, Japan). Autoradiographic images (pixel size ¼ 50 Â 50 mm 2 ) were processed and quantified using the software Fiji (Schindelin et al., 2012). Image pixel values were converted into photostimulated luminescence (PSL)/mm 2 units. Plastic tritium standards (Amersham, Uppsala, Sweden) were used for transformation of the image intensity values into radioactivity units and binding density, expressed as pmol/g tissue. Images selected for figure illustrations were transformed into pmol/g tissue units and were spatially normalized by rotation and translation for visual comparison of radioligand binding in adjacent sections.
The inhibition constant corresponding to the concentration of test compounds required for half-maximum inhibition of radioligand binding (IC 50 ) was estimated by nonlinear parametric curve fitting using GraphPad Prism 7.0 (GraphPad Software, Inc.). The equilibrium dissociation constant of the competitive ligand (K i ) was calculated from the IC 50 using the Cheng-Prusoff equation (Cheng and Prusoff, 1973) based on the equilibrium dissociation constants previously reported for the radioligands (Table S1). Correlations between the regional binding densities for the radioligands were assessed using Pearson correlation coefficient.
Images of the average brain radioactivity and time curves for regional radioactivity after administration of [ 11 C]ABP688 and [ 11 C] AZD9272 in an individual NHP are presented in Fig. 1 and Supplementary Fig. S1, respectively. The binding of [ 11 C]ABP688 was highest in the prefrontal and temporal cortices and striatum, and lower in the midbrain and cerebellum. A different regional distribution of radioactivity was observed for [ 11 C]AZD9272, which displayed high binding in the midbrain, thalamus and ventral striatum relative to the binding in cortical regions. Radioactivity concentration for [ 11 C]AZD9272 in the cerebellum was in the range of that for cortical regions and higher than for the occipital cortex (Fig. S1).
For both tracers the Logan linear graphical analysis yielded a linear phase from 39 min in all of the regions analyzed. The V T values obtained by this method (Table 1 and Table S2) confirmed differences in the rank order of regional availability of binding sites for [ 11 C]ABP688 (CAU~VST~ACC > INS > pTC > HIP > PUT > dPFCT HA > OC > VM > CER) and [ 11 C]AZD9272 (VST > CAU > THA~ACC > HIP > VM~INS~PUT > pTC~dPFC~CER > OC). Slightly higher V T values for [ 11 C]AZD9272 than for [ 11 C]ABP688 were observed in cortical regions, whereas corresponding values were markedly (about two-fold) higher in midbrain, cerebellum, ventral striatum and thalamus.
The binding of [ 11 C]AZD9272 was, to a major extent (93% and 98%, respectively), inhibited by AZD9272 (0.4 mg/kg) or fenobam (2.0 mg/kg; Fig. 2 and Fig. S2), but was only partially inhibited by ABP688 or MTEP at the doses administered (0.25 mg/kg-2.0 mg/ kg). The occupancy induced by ABP688 at all the three doses ranged between 46 and 48% and was independent of dose. After administration of MTEP occupancy was approx. 3% at the 0.25 mg/kg dose and 20% at the highest (2.0 mg/kg) dose level. However, when using [ 11 C]ABP688 as a radioligand the occupancy induced by AZD9272 (0.4 mg/kg) was close to 100% (Fig. 2).

Autoradiography studies
Autoradiography images provided high-resolution anatomical information regarding the regional localization of [ 3 H]AZD9272 binding sites in comparison to the standard mGluR5 radioligand [ 3 H]M-MTEP ( Fig. 3 and Fig. S3). In accordance with PET studies using [ 11 C]AZD9272 as a radioligand,  (Table S1). The regional binding of [ 3 H] Fig. 1. Fused MR and PET images showing distribution of brain radioactivity after injection of [ 11 C]ABP688 or [ 11 C]AZD9272 in the same non-human primate examined using both radioligands. The images represent average radioactivity from 3 to 93 min after injection. Image intensity is displayed as standardized uptake value (SUV), corresponding to radioactivity concentration normalized for injected radioactivity and body weight. ABP688 was correlated, at a statistically significant level (Pearson's r ¼ 0.87; P ¼ 0.0002), with binding for [ 3 H]M-MTEP, but not with that for [ 3 H]AZD9272 (Pearson's r ¼ À0.012; P ¼ 0.97; Table S3).
High binding of [ 3 H]AZD9272 relative to that of [ 3 H]M-MTEP was observed in midbrain structures, including central gray matter, substantia nigra, and the inferior colliculus ( Fig. 3 and Fig. S3). In the thalamus preferential binding of [ 3 H]AZD9272 was observed in mediodorsal and laterodorsal nuclei (Fig. 3) Fig. S3).
In autoradiographic competition binding studies carried out using [ 3 H]M-MTEP, unlabeled ABP688 or AZD2066 potently inhibited the binding (K i 0.35 nM and 3.4 nM, respectively), whereas fenobam displayed significantly lower affinity (K i ¼ 97 nM; Fig. 4A). Conversely, the binding of [ 3 H]AZD9272 was Table 1 Comparison of regional estimates of the total (V T ) and specific (V S ) volume of distribution for [ 11 C]ABP688 (n ¼ 3) and [ 11 C]AZD9272 (n ¼ 4) determined using PET studies in non-human primate brain. Mean and range of V T and V S , and ratios of mean regional V S values for [ 11 C]AZD9272 and [ 11 C]ABP688 are presented. Individual estimates of V T for the two radioligands are provided in Supplementary Table S2 2. PET studies of drug-induced occupancy calculated using a graphical analysis method (Cunningham et al., 2010). Occupancy of AZD9272, fenobam, ABP688 and MTEP using [ 11 C]AZD9272 as a radioligand (AeD) and of AZD9272 using [ 11 C]ABP688 as a radioligand (E). (F) Bound radioligand, relative to the binding at baseline, as a function of dose of the test compounds.

Discussion
AZD9272 represents a distinctive class of mGluR5 selective compounds and its structural difference from other described mGluR5 ligands is of considerable interest. In order to characterize its binding properties, in vivo and in vitro imaging studies using carbon-11-and tritium-labeled analogues of AZD9272 ([ 11 C] AZD9272 and [ 3 H]AZD9272, respectively) were carried out in the NHP brain. Specificity of the binding of radiolabeled AZD9272 towards the mGluR5 was assessed by comparing its regional distribution pattern with that of previously described mGluR5 radioligands and also by competition binding studies, in the absence and presence of unlabeled mGluR5 compounds. The two major observations of the study were that the regional distribution pattern for radiolabeled AZD9272 in NHP brain differs significantly from that of previously described mGluR5 radioligands, and that AZD9272 cannot be fully displaced from its binding sites by MTEP and ABP688 at concentrations predicted to saturate the mGluR5. These observations imply that AZD9272 recognizes additional binding sites in NHP brain that are not shared by MPEP-related mGluR5 compounds.
The binding of the radioligands [ 11 C]ABP688 and [ 3 H]M-MTEP in the NHP brain was high in cortical and striatal regions, and lower in brainstem and cerebellum. However, the regional distribution pattern was different for radiolabeled AZD9272 with predominant binding in midbrain, cerebellum, thalamus and ventral striatum.   ABP688 (Ametamey et al., 2007;DeLorenzo et al., 2011), [ 18 F]FPEB (Sullivan et al., 2013) and other MPEP-related mGluR5 radioligands (Lohith et al., 2017;Pillai and Tipre, 2016), indicating that the additional component of AZD9272 binding is not shared with radioligands belonging to the MPEP series. Moreover, the observation that the unique AZD9272 binding sites, not recognized by classical mGluR5 radioligands, are widely distributed in primate brain may be of relevance for interpretation of the pharmacology of AZD9272 in humans.
In the PET studies using [ 11 C]AZD9272 as a radioligand the binding was almost completely inhibited by fenobam (2.0 mg/kg). In contrast, unlabeled MTEP and ABP688 inhibited [ 11 C]AZD9272 binding only partially, by 20% and 46%, respectively, at dosing conditions predicted to occupy 90% of the brain mGluR5 sites (DeLorenzo et al., 2011;Mathews et al., 2014). These observations were corroborated by the autoradiography findings of a distinct binding pattern of [ 3 H]AZD9272 characterized by high displacement by fenobam, compared to other mGluR5 ligands evaluated. In addition fenobam displayed over 10-fold higher affinity for inhibiting the binding of [ 3 H]AZD9272 (K i ¼ 6 nM) than that of [ 3 H]M-MTEP (K i ¼ 97 nM). The latter observation is consistent with the moderate affinity for inhibiting the binding of [ 3 H]MPEP at human mGluR5 previously reported for fenobam (K i ¼ 53 nM; Porter et al., 2005). These findings provide significant evidence for previously unknown binding sites that are recognized by AZD9272 and fenobam, but are not shared by the other mGluR5 ligands evaluated.
The different competition binding profile and regional distribution pattern in comparison to the MPEP-like radioligands could suggest that AZD9272 is not specific for the mGluR5. However, recognition of targets other than the mGluR5 is not supported by initial screening studies in cell lines expressing the human mGluR5, where AZD9272 has been found to be highly selective towards mGluR5 over other molecular targets . Indeed, the selectivity in vitro and the observation that the binding can be inhibited by the structurally dissimilar mGluR5 compound fenobam, and to some extent, by MPEP-related mGluR5 compounds, suggest that the observed AZD9272 binding is specific to the mGluR5. With regard to fenobam, this compound has been characterized as an mGluR5 NAM displaying a high degree of selectivity over other biomolecules in vitro, with exception of the adenosine A 3 receptor (IC 50 ca. 6 mM; Porter et al., 2005). Based on the low levels of the A 3 receptor in brain (Rivkees et al., 2000) and low affinity of AZD9272 for this receptor  recognition of the A 3 receptor is an unlikely explanation for the residual signal in imaging studies using radiolabeled AZD9272. Nevertheless, binding to other, non-mGluR5-related sites that could be shared by fenobam and AZD9272 cannot be completely excluded.
As an alternative explanation, evidence for the existence of multiple ligand recognition sites at the mGluR5 protein makes it possible for AZD9272 and MPEP-related compounds to recognize different sites of the mGluR5. Indeed, several separate binding sites have been identified in the mGluR5 modulator binding pocket (Chen et al., 2008;Hammond et al., 2010;Noetzel et al., 2013;O'Brien et al., 2004), suggesting possibility of non-competitive interaction between AZD9272 and MPEP-like compounds at the receptor (Rook et al., 2015). However, the binding of [ 3 H]AZD9272 at human mGluR5 in vitro has been reported to be inhibited, in a competitive manner, by MPEP, supporting binding at a common site . Moreover, this hypothesis would not account for the differences in regional distribution patterns between radiolabeled AZD9272 and MPEP-like radioligands. In theory, regionally distinct binding patterns could be explained by differential selectivity of mGluR5 NAMs for splice variants of the receptor located in different brain regions (Malherbe et al., 2002;Minakami et al., 1994), but this hypothesis seems also unlikely given the similar ligand binding properties previously reported at mGluR5 splice variants (Malherbe et al., 2002).
Another explanation that would account for observations of a distinct regional distribution pattern of radiolabeled AZD9272 can be based on the prevailing view that the mGluR5 forms heteromeric complexes with other signaling proteins, a biological condition that may alter ligand pharmacology (Fuxe et al., 2009). It could thus be hypothesized that AZD9272 binds with higher affinity to mGluR5 heteromeric complexes, with different anatomical localization than the homomeric or monomeric forms of the receptor. Of the identified candidate heteromeric complex partners the m opioid receptor (Schr€ oder et al., 2009) is of particular interest since its regional brain localization resembles the distinctive parts of the binding pattern of AZD9272 in primate brain (Daunais et al., 2001). Further studies are required to assess the potential impact of receptor heteromerization on the binding affinity of AZD9272, fenobam and the MPEP-like mGluR5 ligands.
CNS adverse events, involving psychosis-like phenomena, have been reported in early clinical trials of fenobam (Friedmann et al., 1980;Pecknold et al., 1982). Psychotomimetic effects, similar to those reported for fenobam, have also been observed during the clinical development of AZD9272 (Ståhle et al., 2012). The present findings of common binding sites for AZD9272 and fenobam in primate brain may provide insights on molecular targets associated with such psychotomimetic properties. Along these lines it is of interest to note that the residual component of the AZD9272 binding was identified in several brain regions implicated in the treatment of schizophrenia (Clinton and Meador-Woodruff, 2004;Howes et al., 2017;Mikell et al., 2016), such as the ventral striatum, thalamus and midbrain. Based on these observations, [ 11 C]AZD9272 may represent an interesting tool for research into the pathophysiology of schizophrenia and maybe also for identification of novel targets for antipsychotic treatment.
A limitation of the present study is the small number of NHPs examined per pre-treatment condition. Assessment of fenobam occupancy in vivo was based on a single experiment conducted at high drug exposure, thus precluding the evaluation of dose dependency and drug affinity in vivo. However, findings from our in vitro studies support concentration-dependent inhibition of [ 3 H] AZD9272 binding by fenobam, with fenobam displaying higher affinity to [ 3 H]AZD9272 binding sites than other mGluR5 NAMs evaluated. Nevertheless, additional PET studies employing multiple doses of fenobam and plasma measurements of drug concentration to determine affinity in vivo are warranted to further assess the binding of fenobam at AZD9272 binding sites.
The present observations in NHPs differ from previous findings of autoradiography studies in rat brain sections , where the binding of [ 3 H]AZD9272 has been reported to display a regional distribution pattern and ligand competition properties consistent with that of MPEP-like radioligands . These discrepancies suggest that the residual AZD9272 binding component may be unique to primates. Indeed, observations from initial PET studies using [ 11 C]AZD9272 and a low resolution PET system in human subjects  suggest a distribution pattern consistent with that of NHP brain. Future investigations using radiolabeled AZD9272 and a high resolution (HRRT) PET system would allow the mapping of binding sites for AZD9272 in human brain at higher anatomical resolution.
In conclusion, our findings support that AZD9272 binds to a population of sites in the NHP brain that are also recognized by the pro-psychotic mGluR5 compound fenobam. This binding can only be partially displaced by MPEP-like compounds including MTEP and ABP688. The distinct binding of AZD9272 in thalamus, ventral striatum and midbrain may suggest relevance of this additional binding component to the pathophysiology of psychiatric disorders involving mGluR5 and therapies targeting this receptor.

Declaration of interest
The studies were supported by AstraZeneca Pharmaceuticals. Dr. Jur eus and Dr. Raboisson are past employees and Dr. Johnstr€ om and Prof. Farde current employees of AstraZeneca Pharmaceuticals. Prof. Farde has served as a panel member for evaluation of the Research Programs of the Faculty of Medicine, University of Helsinki, Finland. Dr. Finnema is a current employee of AbbVie Inc. The other authors declare no potential conflicts of interest.